$git clone https://github.com/cloudshipai/stationBuild, test, and deploy intelligent agent teams. Self-hosted. Git-backed. Production-ready.
| 1 |  |
| 2 | |
| 3 | # Station - AI Agent Orchestration Platform |
| 4 | |
| 5 | [](./TESTING_PROGRESS.md) [](./.github/workflows/ci.yml) |
| 6 | |
| 7 | **Build, test, and deploy intelligent agent teams. Self-hosted. Git-backed. Production-ready.** |
| 8 | |
| 9 | [Quick Start](#quick-start) | [Real Example](#real-example-sre-incident-response-team) | [Deploy](#deploy-to-production) | [Documentation](https://docs.cloudshipai.com) |
| 10 | |
| 11 | --- |
| 12 | |
| 13 | ## Why Station? |
| 14 | |
| 15 | Build multi-agent systems that coordinate like real teams. Test with realistic scenarios. Deploy on your infrastructure. |
| 16 | |
| 17 | **Station gives you:** |
| 18 | - ✅ **Multi-Agent Teams** - Coordinate specialist agents under orchestrators |
| 19 | - ✅ **Built-in Evaluation** - LLM-as-judge tests every agent automatically |
| 20 | - ✅ **Git-Backed Workflow** - Version control agents like code |
| 21 | - ✅ **One-Command Deploy** - Push to production with `stn deploy` |
| 22 | - ✅ **Full Observability** - Jaeger traces for every execution |
| 23 | - ✅ **Self-Hosted** - Your data, your infrastructure, your control |
| 24 | |
| 25 | --- |
| 26 | |
| 27 | ## Quick Start (2 minutes) |
| 28 | |
| 29 | ### Prerequisites |
| 30 | |
| 31 | - **Docker** - Required for Jaeger (traces and observability) |
| 32 | - **AI Provider** - Choose one: |
| 33 | - **CloudShip AI** (Recommended) - `STN_CLOUDSHIP_KEY` or `CLOUDSHIPAI_REGISTRATION_KEY` |
| 34 | - `OPENAI_API_KEY` - OpenAI (gpt-5-mini, gpt-5, etc.) |
| 35 | - `GEMINI_API_KEY` - Google Gemini |
| 36 | - `ANTHROPIC_API_KEY` - Anthropic (claude-sonnet-4-20250514, etc.) |
| 37 | |
| 38 | ### 1. Install Station |
| 39 | |
| 40 | ```bash |
| 41 | curl -fsSL https://raw.githubusercontent.com/cloudshipai/station/main/install.sh | bash |
| 42 | ``` |
| 43 | |
| 44 | ### 2. Initialize Station |
| 45 | |
| 46 | Choose your AI provider: |
| 47 | |
| 48 | <details open> |
| 49 | <summary><b>CloudShip AI (Recommended)</b></summary> |
| 50 | |
| 51 | Use CloudShip AI for optimized inference with Llama and Qwen models. This is the default when a registration key is available. |
| 52 | |
| 53 | ```bash |
| 54 | # Set your CloudShip registration key |
| 55 | export CLOUDSHIPAI_REGISTRATION_KEY="csk-..." |
| 56 | # Or use: export STN_CLOUDSHIP_KEY="csk-..." |
| 57 | |
| 58 | stn init --provider cloudshipai --ship # defaults to cloudship/llama-3.1-70b |
| 59 | ``` |
| 60 | |
| 61 | **Available models:** |
| 62 | - `cloudship/llama-3.1-70b` (default) - Best balance of performance and cost |
| 63 | - `cloudship/llama-3.1-8b` - Faster, lower cost |
| 64 | - `cloudship/qwen-72b` - Alternative large model |
| 65 | |
| 66 | </details> |
| 67 | |
| 68 | <details> |
| 69 | <summary><b>Claude Max/Pro Subscription (⚠️ DEPRECATED)</b></summary> |
| 70 | |
| 71 | > **⚠️ DEPRECATED: Anthropic OAuth is currently unavailable.** |
| 72 | > |
| 73 | > Anthropic has restricted third-party use of OAuth tokens. This authentication method is not working until further notice. |
| 74 | > |
| 75 | > **Please use one of the following alternatives:** |
| 76 | > - **OpenAI API Key** (recommended) |
| 77 | > - **Google Gemini API Key** |
| 78 | > - **Anthropic API Key** (pay-per-token, not subscription-based) |
| 79 | |
| 80 | ~~Use your existing Claude Max or Claude Pro subscription - no API billing required.~~ |
| 81 | |
| 82 | ```bash |
| 83 | # ❌ NOT WORKING - Anthropic OAuth disabled |
| 84 | # stn init --provider anthropic --ship |
| 85 | # stn auth anthropic login |
| 86 | ``` |
| 87 | |
| 88 | </details> |
| 89 | |
| 90 | <details> |
| 91 | <summary><b>OpenAI (API Key)</b></summary> |
| 92 | |
| 93 | ```bash |
| 94 | export OPENAI_API_KEY="sk-..." |
| 95 | stn init --provider openai --ship # defaults to gpt-5-mini |
| 96 | ``` |
| 97 | |
| 98 | </details> |
| 99 | |
| 100 | <details> |
| 101 | <summary><b>Google Gemini (API Key)</b></summary> |
| 102 | |
| 103 | ```bash |
| 104 | export GEMINI_API_KEY="..." |
| 105 | stn init --provider gemini --ship |
| 106 | ``` |
| 107 | |
| 108 | </details> |
| 109 | |
| 110 | This sets up: |
| 111 | - ✅ Your chosen AI provider |
| 112 | - ✅ [Ship CLI](https://github.com/cloudshipai/ship) for filesystem MCP tools |
| 113 | - ✅ Configuration at `~/.config/station/config.yaml` |
| 114 | |
| 115 | ### 3. Start Jaeger (Tracing) |
| 116 | |
| 117 | Start the Jaeger tracing backend for observability: |
| 118 | |
| 119 | ```bash |
| 120 | stn jaeger up |
| 121 | ``` |
| 122 | |
| 123 | This starts Jaeger UI at [http://localhost:16686](http://localhost:16686) for viewing agent execution traces. |
| 124 | |
| 125 | ### 4. Connect Your MCP Client |
| 126 | |
| 127 | Choose your editor and add Station: |
| 128 | |
| 129 | <details> |
| 130 | <summary><b>Claude Code CLI</b></summary> |
| 131 | |
| 132 | ```bash |
| 133 | claude mcp add station -e OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318 --scope user -- stn stdio |
| 134 | ``` |
| 135 | |
| 136 | Verify wi |